Research On Active Lane Change System Based On Driving Intention Recognition

With the improvement of life,people focus more and more on the comfort and safety during driving.Since it could liberate the driver from tedious driving,the research on autonomous driving become more and more popular.But due to restrictions in policies and regulations,it is still a long time before the arrival of the era of automatic driving.As a transitional solution between the traditional vehicle and autonomous vehicle,the driver assistance system,such as lane keeping system and adaptive cruise control system,receives more and more attention.Besides,it is very important for driver assistance system to recognize the driver intention accurately.The driver often need to change lanes during a long journey.In this context,this paper conducts research on driving intention recognition and applies the recognition results in active lane change.The Gaussian Mixture Hidden Markov Model is used to recognize the driving intention.The lane change processes are simulated in Matlab/Carsim software with a speed of 65km/h.There are 2 schemes designed to train the lane change model.The first one takes the steer wheel angle and steer wheel speed as observation variables,and uses the data from the driver starts to take lane change action to the steer angle arrives its maximum.The second one takes the steer wheel angle,the angular speed and the yaw angle as observation variables,and uses the data through the whole lane change process.As for lane keep model,data are obtained from driving simulation test bench.The verification of the model and the analysis of the recognition ability are conducted for each scheme with different sliding-time-window after the models are built.Considering the differences of the signal characteristics under different vehicle speed,the data of the lane change under 100 km/h are used to verify the model.The result shows that both of those schemes could recognize the lane change operation with a speed of 100kn/h accurately,too.In this thesis,the recognition delay time,the recognition advance time,the lasted recognition time and the position of the gravity center of the vehicle at the end of the correct recognition are defined as the evaluation indexes,which could reflect the recognition effect of the whole lane change process.As a result,both schemes could recognize the lane change operation within 600 ms after it begins.And the recognitiontime is about 1.9s before the vehicle cross the lane line.Besides,the second schemes could recognize the whole lane change process.Based on the driver intention recognition and the model predictive control,a active lane change system is designed.This system could plan,and then track the lane change trajectory after it recognizes the lane change operation.The system is firstly verified by simulation,during which the steer angle is not smooth if the path tracking controller intervenes immediately after the lane change intention is recognized,and a solution is found by delay the time the controller intervenes.Finally,the system is tested in driving simulation test bench with a driver who conducts lane change behavior.The result shows that this system could be put into practice,only if the transition problem between the driver and the controller is well handled.